University of Windsor Scholarship at UWindsor OSSA Conference Archive OSSA 4 May 17th, 9:00 AM - May 19th, 5:00 PM Anecdotal Reasoning Louis F. Groarke York University Follow this and additional works at: https://scholar.uwindsor.ca/ossaarchive Part of the Philosophy Commons Groarke, Louis F., "Anecdotal Reasoning" (2001). OSSA Conference Archive. 52. https://scholar.uwindsor.ca/ossaarchive/OSSA4/papersandcommentaries/52 This Paper is brought to you for free and open access by the Conferences and Conference Proceedings at Scholarship at UWindsor. It has been accepted for inclusion in OSSA Conference Archive by an authorized conference organizer of Scholarship at UWindsor. For more information, please contact [email protected]. Title: Anecdotal Reasoning Author: Louis Groarke © 2001 Louis Groarke 1. Introduction In this presentation, I want to argue that anecdotal reasoning is not inherently vicious. The widespread condemnation of anecdotal reasoning one finds in the informal logic literature is based on an analysis of inductive generalizations which is ultimately flawed. As Aristotle explains, induction, is a form of insight, not simply a case of mechanical enumeration. As such, it is a more complex argument form than is often recognized. The following is only intended as an introductory account of the issue. 2. Hasty Generalization The argument form I want to examine is usually referred to, in the pedagogical literature as “hasty generalization” (Morris, 43; Toulmin, 151-155, 173; Kelly, 266-267; Groarke, Tindale, Fisher, 219; Conway, Munson, 129). Authors also call this allegedly fallacious reasoning strategy: “hasty inductive generalization”(Govier, 344), “appeal to anecdotal evidence” (Moore, Parker, 396-97), “glittering generality” (Woods, Walton, 65), the fallacy of “insufficient sample” (Damer, 109-110), the “fallacy of insufficient statistics” (Salmon, 56), the “fallacy of generalization from too few cases” (Fearnside, Holther, 13), or, in a reference to traditional categories, “secundum quid (Hamblin, 28-29, 46-47; Woods, Walton, 65). On the standard explanation, hasty generalization is a matter of “jumping to conclusions” (Groarke, Tindale, Fisher, 219). We commit a hasty generalization, when we “generalize too quickly, on the basis of insufficient evidence” (Kelly, p. 266). This results in a sweeping claim which has been drawn “from too small a sample” (Damer, p. 109). As Govier explains, every- day generalizations are often “based on an exceedingly small number sample of cases— sometimes only one or two. A person rather carelessly assumes that the case or cases that have come to her attention are more than just episodes or isolated events; she [mistakenly] assumes, without sufficient warrant, that they indicate a general tendency or trend” (Govier, 344). Engels offers the following example of hasty generalization: “The welfare program is totally unnecessary. Why, I know a guy who runs a very lucrative gambling operation and who drives his new Cadillac downtown every week to collect is welfare check” (Engels, 43-44). This appeal to anecdote is a conspicuous example of biased reasoning. The speaker assumes, on the basis of single exceptional example, that everyone on welfare is cheating the system. Obviously, one cannot make a valid generalization about all the people on welfare on the basis of a single atypical instance. So clearly, this is a mistake in reasoning. Contemporary textbook authors often use examples of racist argument or social stereotyping to make the same point. Kelly typically warns, “ A single bad experience while traveling can prejudice our view of an entire city or country. Most of us have stereotypes about ethnic groups, professions, or people from different regions of the country based on our exposure to a few individuals. Even a judgement about the character or personality of a particular individual is a generalization drawn from our observation of that individual on specific occasions” (Kelly, 1 Anecdotal Reasoning 2 266). If, however, racist stereotyping is logically problematic, I will argue that so-called “hasty generalization” is not always a fallacious argument strategy. 3. Two-Mistakes-in-One Contemporary authors sometimes distinguish between two kinds of hasty generalization. Toulmin writes that “we commit fallacies of hasty generalization when we: (1) draw a general conclusion from too few specific instances...or, alternatively, when we (2) draw a general conclusion from atypical examples” (Toulmin, 151). But these do not seem to be separate mistakes. If the burden of proof is on authors to demonstrate that they have provided a correct instance of reasoning, then one mistake entails the other. Consider. Suppose I commit mistake (1): I investigate “too few instances” of something (Toulmin, 173). If I investigate too few instances, then I cannot know whether the particular instances I investigate are typical or typical. This leaves open the possibility that these instances are atypical. This is why this reasoning strategy is considered to be fallacious. If I look at too few instances, then I cannot know if I am investigating atypical instances, and if I am basing my conclusions on atypical examples, I will be lead to false conclusions. So mistake (1) presupposes, in some strong sense, the possibility of mistake (2). Suppose, however, I commit mistake (2): I investigate “exceptional rather than typical cases of something” (Toulmin, 173). A minimally rational individual would not base sweeping conclusions about a whole class of entities on the basis of atypical instances. If I am confusing atypical and typical cases, this can only be because I have not looked at enough cases to realize that the cases I am investigating are atypical. So mistake (2) presupposes mistake (1). When a reasoner generalizes from atypical cases, this is because of inadequate evidence. We may in hindsight ascertain the atypical nature of examples studied, but this is only because we have studied further examples that the original reasoner did not, for one reason or another, consult. If contemporary authors distinguish between reasoning from too few examples and reasoning from atypical examples, Woods and Walton, in a discussion of sampling methods, draw an analogous distinction between the fallacy of “insufficient statistics” and the fallacy of “biased statistics” (Woods, Walton, 66). This is a useful statistical distinction between the number of cases studied and the method employed for selecting them. Nonetheless, one could argue that these statistical problems are inter-related. When we study two few cases, this may result in a sample made up of atypical cases. And when we use a biased selection procedure, this is due to a lack of familiarity with the object and population under the study. If researchers were familiar with enough examples, they could presumably devise an accurate sampling method. Although there may be cases that are so complex that the practical matter of sample-selection is problematic (the necessary knowledge just isn’t available), they need not detain us here. For simplicity sake, consider the fallacy of hasty generalization to be the practice of reasoning from very few cases to some kind of inductive generalization. This may, of course, involve generalizing from only one or two examples. I mean to show that many cases of so- called hasty generalization are clearly not fallacious. Anecdotal Reasoning 3 4. Anecdotal Evidence Reasoning from anecdotal evidence is usually classified as a form of hasty generalization. In its literary meaning, the term anecdote refers to “a short, entertaining account of some happening, usually personal or biographical” (Websters). We are interested in the logical status of arguments based on such personal evidence. Define anecdotal reasoning as reasoning from particular experience. Anecdotal reasoning is not necessarily unsystematic but it is not, in any strict sense, statistical. Naturalists may study the animal kingdom in a highly systematic way, but their investigations revolve around personal experience. Again the literary historian may base a systematic account of history on the personal experience of diverse historical figures. These might count as unusually cogent forms of anecdotal reasoning. But such reasoning is generally overlooked in the informal logic literature. In a discussion of hasty generalization, Moore and Parker cite a statistical rule of thumb: “anecdotes prove nothing” (Moore, Parker, 396). As they explain: “An anecdote that shows one or two Xs have feature f proves only that one or two Xs have feature f.” It “doesn’t prove anything about Xs in general” (Moore, Parker, 396-97). This is the standard view. But the kinds of examples invoked in support of this generalization are cases of special pleading. They have been carefully selected so as to support a partisan hypothesis. There are cases of reasoning from one or two examples that are clearly cogent. Begin with a silly, but useful example. Suppose I encounter a circle. That is, I have one particular experience with a circle. Perhaps I see the shape in a picture in the newspaper. I inspect the shape and come to the conclusion: the circumference of every circle is curvilinear. Clearly, this is a true conclusion. If, however, the conclusion is true, I have inferred it from a careful consideration from a single example. I might tell my buddies at the beach, “Hey, I saw this circle,” and drawing a crude shape in the sand, I might tell them: “See, it was something like this; the sides are all curvilinear.” And they might be lead, correctly, to accept the true conclusion that all circles have sides that are curvilinear. Again, suppose I see one adult elephant. I conclude that adult elephants have trunks, elongated, pliable snouts that they can fold, pick up things with, spray water through. I don’t just randomly guess that all adult elephants have trunks; I carefully observe the creature and apply perhaps some ideas about biology and species differentiation to my observation. This is an accurate, if hasty, generalization. And clearly, it can be inferred on the basis of one example.